Instructions to use AIPROD-C2/scriptmind-v1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use AIPROD-C2/scriptmind-v1 with Transformers:
# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("AIPROD-C2/scriptmind-v1", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Update training_script.py
Browse files- training_script.py +12 -15
training_script.py
CHANGED
|
@@ -2,50 +2,47 @@ from transformers import AutoTokenizer, AutoModelForCausalLM, Trainer, TrainingA
|
|
| 2 |
from datasets import load_dataset
|
| 3 |
import torch
|
| 4 |
|
| 5 |
-
# Charger TES données
|
| 6 |
print("📥 Loading AIPROD script dataset...")
|
| 7 |
-
dataset = load_dataset("AIPROD-C2/aiprod-script-dataset"
|
| 8 |
|
| 9 |
-
#
|
| 10 |
model_name = "distilgpt2"
|
| 11 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
| 12 |
model = AutoModelForCausalLM.from_pretrained(model_name)
|
| 13 |
|
| 14 |
-
# Ajouter
|
| 15 |
tokenizer.pad_token = tokenizer.eos_token
|
| 16 |
|
| 17 |
-
#
|
| 18 |
def tokenize_function(examples):
|
| 19 |
return tokenizer(examples["text"], truncation=True, padding="max_length", max_length=128)
|
| 20 |
|
| 21 |
-
print("🔧 Tokenizing dataset...")
|
| 22 |
tokenized_datasets = dataset.map(tokenize_function, batched=True)
|
| 23 |
|
| 24 |
-
#
|
| 25 |
training_args = TrainingArguments(
|
| 26 |
-
output_dir="./scriptmind-v1",
|
| 27 |
-
|
| 28 |
-
num_train_epochs=3,
|
| 29 |
per_device_train_batch_size=4,
|
| 30 |
save_steps=500,
|
| 31 |
save_total_limit=2,
|
| 32 |
-
prediction_loss_only=True,
|
| 33 |
logging_dir='./logs',
|
|
|
|
| 34 |
)
|
| 35 |
|
| 36 |
-
# Entraîneur
|
| 37 |
trainer = Trainer(
|
| 38 |
model=model,
|
| 39 |
args=training_args,
|
| 40 |
train_dataset=tokenized_datasets["train"],
|
| 41 |
)
|
| 42 |
|
| 43 |
-
#
|
| 44 |
print("🚀 Starting AIPROD scriptmind-v1 training...")
|
| 45 |
trainer.train()
|
| 46 |
|
| 47 |
-
#
|
| 48 |
trainer.save_model("./scriptmind-v1-final")
|
| 49 |
tokenizer.save_pretrained("./scriptmind-v1-final")
|
| 50 |
-
|
| 51 |
print("✅ scriptmind-v1 training completed!")
|
|
|
|
| 2 |
from datasets import load_dataset
|
| 3 |
import torch
|
| 4 |
|
| 5 |
+
# Charger TES données AIPROD
|
| 6 |
print("📥 Loading AIPROD script dataset...")
|
| 7 |
+
dataset = load_dataset("AIPROD-C2/aiprod-script-dataset")
|
| 8 |
|
| 9 |
+
# Configuration
|
| 10 |
model_name = "distilgpt2"
|
| 11 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
| 12 |
model = AutoModelForCausalLM.from_pretrained(model_name)
|
| 13 |
|
| 14 |
+
# Ajouter le token de padding
|
| 15 |
tokenizer.pad_token = tokenizer.eos_token
|
| 16 |
|
| 17 |
+
# Préparation des données
|
| 18 |
def tokenize_function(examples):
|
| 19 |
return tokenizer(examples["text"], truncation=True, padding="max_length", max_length=128)
|
| 20 |
|
| 21 |
+
print("🔧 Tokenizing AIPROD dataset...")
|
| 22 |
tokenized_datasets = dataset.map(tokenize_function, batched=True)
|
| 23 |
|
| 24 |
+
# Entraînement
|
| 25 |
training_args = TrainingArguments(
|
| 26 |
+
output_dir="./scriptmind-v1-output",
|
| 27 |
+
num_train_epochs=5,
|
|
|
|
| 28 |
per_device_train_batch_size=4,
|
| 29 |
save_steps=500,
|
| 30 |
save_total_limit=2,
|
|
|
|
| 31 |
logging_dir='./logs',
|
| 32 |
+
logging_steps=50,
|
| 33 |
)
|
| 34 |
|
|
|
|
| 35 |
trainer = Trainer(
|
| 36 |
model=model,
|
| 37 |
args=training_args,
|
| 38 |
train_dataset=tokenized_datasets["train"],
|
| 39 |
)
|
| 40 |
|
| 41 |
+
# LANCEMENT !
|
| 42 |
print("🚀 Starting AIPROD scriptmind-v1 training...")
|
| 43 |
trainer.train()
|
| 44 |
|
| 45 |
+
# Sauvegarde finale
|
| 46 |
trainer.save_model("./scriptmind-v1-final")
|
| 47 |
tokenizer.save_pretrained("./scriptmind-v1-final")
|
|
|
|
| 48 |
print("✅ scriptmind-v1 training completed!")
|